1. Technical Field
The present invention relates to an image processing device for obtaining a processing result through a process using a set of parameters defined in advance on image data obtained by imaging a measuring target object.
2. Related Art
In the field of FA (Factory Automation) and the like, signal processing and the like are performed on various types of data acquired with use of various types of sensors from a measuring target object such as a half-completed product in the manufacturing process or a product before shipment, and tests, measurements, discrimination and the like on the measuring target object are often carried out based on the signal processed data. Typically, there is widely used an image processing device for acquiring image data by imaging the measuring target object with an imaging device or the like, and optically testing defects in the measuring target object, optically measuring the size of a specific portion of the measuring target object, and detecting a character string appearing in the measuring target object based on the acquired image data (see e.g., Japanese Unexamined Patent Publication No. 08-101139).
Japanese Unexamined Patent Publication No. 2002-008013, for example, discloses an outer appearance test program used in an outer appearance testing device for determining defects of the outer appearance of an article using image data. More specifically, Japanese Unexamined Patent Publication No. 2002-008013 has an object to perform an accurate and highly reliable test of an outer appearance.
Japanese Unexamined Patent Publication No. 2006-085616 discloses an image processing algorithm evaluation method using an evaluation value reflecting a separation capacity between an extraction target site and a non-extraction target side in an image. More specifically, disclosed is a technique of performing a performance evaluation of the image processing algorithm for extracting only an extraction target portion by quantifying the degree of separation on a characteristic amount between the extraction target portion and the non-extraction target portion.
When installing such an image processing device on a production line or the like, optimization of a set of parameters related to image processing, test processing, measurement processing, discrimination processing, and the like is a very troublesome task. In other words, in the image processing device described above, the result on the measuring target object appearing in the image data is judged in a comprehensive manner after performing a similar process on one piece of input image data a plurality of times. Furthermore, the number of parameters related to image processing in the image processing device is relatively large, and the result of image processing sometimes greatly fluctuates only by one parameter being changed. Moreover, the parameters need to be optimized so that a stable result not to cause false determination is obtained by actually performing tests, measurements, discriminations, and the like on the measuring target object.
An actual setting procedure for a set of parameters includes actually performing tests, measurements, discriminations and the like on the measuring target object after setting the set of parameters to a certain set of values, and judging whether or not the set of parameters having been set is appropriate. If judged as inappropriate, the set of parameters is changed to a different set of values, and then tests, measurements, discriminations and the like on the measuring target object are again performed. The setting of the set of parameters, and the evaluation of the processing result based on the actually acquired image data are iteratively repeated until an optimum set of parameters is obtained.
Since such iterative procedures need to be performed, the parameter setting requires a great amount of manpower and many days. Not limited to image data as described above, the parameter setting on various other types of data is also troublesome. The optimum set of parameters is difficult to rapidly obtain by a non-skilled person with sufficient experience.
In the methods described in Japanese Unexamined Patent Publication No. 2002-008013 and Japanese Unexamined Patent Publication No. 2006-085616, there is disclosed the technique of automatically determining a set value related to image processing including necessary parameters according to an evaluation logic defined in advance, but the predefined evaluation logic itself may not necessarily be optimum for all measuring target objects. In other words, since the techniques disclosed in Japanese Unexamined Patent Publication No. 2002-008013 and Japanese Unexamined Patent Publication No. 2006-085616 automatically determine numerous parameters according to an evaluation criteria defined in advance, such techniques may be applied to an image processing device used in a specific purpose, but cannot be applied to a universal (multi-functional) image processing device.